Goal-Talent-Purpose-Process: Managing People and AI With One Playbook
Treat managing agents like managing people: same four levers, different resources.
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 4
- Confidence
- 90%
Zhuo argues management is fundamentally about producing an outcome using available resources, and the same four levers apply whether the resources are people or AI models. You set a clear goal, assemble the right talent (now models with different strengths), align everyone on purpose, and design the process for how it all comes together. This reframes prompting and orchestrating agents as a management discipline rather than a new skill.
Origin
Julie Zhuo's synthesis, extending the management model from her book The Making of a Manager to agentic AI systems.
Core principles
- 01Management is having an outcome and marshalling resources to reach it — the resource used to be people, now it can be models
- 02Different models have different personalities and strengths; you must develop intuition for which tool fits which purpose
- 03Even smarter agents still need the right context and clear high-level instructions from you
- 04The middle lever of traditional management — career development — drops away with agents, but goal/talent/process remain
How to run it
- 1
Define the outcome as your north star
Start from the thing you are trying to get done — a crystal-clear picture of success — before assembling any team or tools.
Pro tip Write down what success looks like specifically enough that you could tell without question whether you hit it.
- 2
Assemble the right mix of talent or models
Choose the people or AI models whose strengths match the job, and get to know each one's personality the way you would a new hire.
Pro tip Also decide which tools the agents get access to — that is part of assembling the team.
Watch out Do not default to one model for everything; strengths differ.
- 3
Align on purpose
Make sure every person and tool knows what they are supposed to do with their talents — the shared goal.
Watch out Different people (and unclear prompts) carry different pictures of success; misalignment is the usual cause of failure.
- 4
Design the process
Decide how the people, models, and tools come together to actually produce the work.
Pro tip Provide rich context and high-level instructions rather than micromanaging every step.
In the wild
Zhuo describes management as figuring out how to use the resources you have to get a thing done — 'used to be people but now it's basically models' — so you assemble the right mix of skills, define the goal, and set the process, identical to running a team.
→ A repeatable mental model that lets experienced managers become effective agent orchestrators without learning a wholly new discipline.
Common mistakes
Treating agent work as a novel skill instead of management
People assume working with AI requires a brand-new capability, missing that the core levers — clear goal, right resource, aligned purpose, defined process — are the same management fundamentals they may already have.
Skipping the specificity of the goal
Because humans think at high levels, they give vague instructions; an agent (like a team) cannot tell whether it has succeeded, so the output disappoints.
Is it for you?
Best for
Managers and builders integrating AI agents into their workflows who already have team-leadership instincts
Not ideal for
Purely mechanical single-shot tasks where no orchestration or goal-setting is needed
From the transcript
“management is just about in my mind having an outcome. So you want to get something done. That's the thing. You have a northstar. You…”
“used to be people but now it's basically models and different models have different strengths”
“these are still the fundamentals of working with agentic systems. Like you still need a goal.”
“it's still very important for us to be able to provide the right context, provide the right highlevel instructions so that we get what we…”
From the episode
From managing people to managing AI: The leadership skills everyone needs now
Julie Zhuo (Facebook VP, Sundial CEO, The Making of a Manage